I've just started to study differential geometry and I've some problems with the first definitions. We have defined a topological manifold with boundary of dimension n as a topological space $M$ such that

  • $M$ is connected
  • $M$ is Hausdorff
  • $M$ is locally homeomorphic to $\mathbb{R}^n_+$

I have to show that

  • $\mathring{B^n}:=\{x\in\mathbb{R^n}:\sum_i x_i^2<1\}$ is a topological manifold with boundary of dimension $n$

The first two conditions are trivial to check. Then let's consider $\Phi:\mathring{B^n}\to\mathbb{R}^n_+$ defined as $\Phi(x_1,\dots,x_n)=(x_1,\dots,e^{x_n})$ which is an homeomorphism.

  • $B^n:=\{x\in\mathbb{R^n}:\sum_i x_i^2\le 1\}$ is a topological manifold with boundary of dimension $n$

As before, it's sufficient to check the last condition. Let's consider

$$ U_1=\{(x_1,\dots,x_n)\in B^n:x_n>-\frac{1}{2}\}\\ U_2=\{(x_1,\dots,x_n)\in B^n:x_n<\frac{1}{2}\} $$ and write $B^n=U_1\cup U_2$. Then I show that $U_i$ are homeomorphic to open subsets of $\mathbb{R}^n_+$ via the maps $$ \phi_1:U_1\to\mathbb{R}^n_+\\ (x_1,\dots,x_n)\mapsto (x_1,\dots,1-x_n) $$ and $$ \phi_2:U_2\to\mathbb{R}^n_+\\ (x_1,\dots,x_n)\mapsto (x_1,\dots,x_n+1) $$

Is it okay?


You didn't say what $\mathbb R^n_+$ is supposed to represent, but from the context, I assume it means $\{(x_1,\dots,x_n): x_n\ge 0\}$.

Your argument for $\mathring{B^n}$ is fine, except for one small mistake: You wrote, "Then let's consider $\Phi:\mathring{B^n}\to\mathbb{R}^n_+$ defined as $\Phi(x_1,\dots,x_n)=(x_1,\dots,e^{x_n})$ which is an homeomorphism." It's not a homeomorphism, because it's not surjective onto $\mathbb R^n_+$; but it's a homeomorphism onto an open subset of $\mathbb{R}^n_+$, which is good enough.

But your argument for $B^n$ has a number of problems. Here are the most important ones:

  • Your two subsets $U_1$ and $U_2$ do not cover all of $B^n$.
  • The images of $\phi_1$ and $\phi_2$ are not open subsets of $\mathbb R^n_+$. For example, $\phi_1(U_1)$ looks like a "bowl full of soup" sitting on the $x^n=0$ hyperplane. More formally, the point $(0,\dots,0) = \phi_1(0,\dots,0,1)$ is in the image of $\phi_1$, but any (relative) neighborhood of the origin in $\mathbb R^n_+$ would have to contain points of the form $(\varepsilon,0,\dots,0)$ for all sufficiently small $\varepsilon$, none of which are in the image of $\phi_1$.

To construct a homeomorphism from an open subset of $B^n$ to an open subset of $\mathbb R^n_+$, you need to find a map that "straightens out" the boundary sphere and maps it into the $x^n=0$ hyperplane. Stereographic projection might be useful for this.

  • $\begingroup$ Could you write explicitly the homeomorphism? $\endgroup$ – avati91 Jul 23 '15 at 9:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.